During the recent years, the demand for low cost multimedia systems that can be incorporated in a computer system or function as a stand alone system has been steadily increasing. There has been some effort to create multimedia processor systems that are primarily software driven or, alternatively, are primarily hardware driven. Some of these multimedia systems employ a separate three dimensional 3D graphic chip that is coupled to a main processor for handling graphics. However, these systems experience delaying bottlenecks during data intensive operations that transfer data between the main processor and the graphics chip.
The bandwidth requirements to implement a 3D graphics system depends on the complexity of the system. Typically, 3D graphic systems include multiple modules in a pipe-lined architecture, such as geometry transformation, lighting transformation, shading or rasterizing by interpolation, texture mapping and texture filtering.
The geometry transformation is the process of transforming the model of a three dimensional object in a three dimensional space into a two dimensional screen space. This process includes the steps of defining the three dimensional model by a plurality of polygons, such as triangles, and transforming these polygons into a two dimensional space.
The geometry of lighting transformation or lighting, is the process of representing the intensity of light reflections from the three dimensional model to a two dimensional screen space.
The texture mapping process provides a mechanism to represent the texture of a three dimensional model. Thus, a texture space is defined in a two dimensional space by two texture coordinates referred to as horizontal “u” coordinate and a vertical “v” coordinate. Each pixel in the texture space is called a texel. Information relating to each texel is stored in an external memory, which can be mapped to the nodes of a corresponding triangle in response to a fetch texel command. The texel color is then blended with the shading color described above resulting in the final color for each node of each triangle. Again a shading by interpolation is employed to find the shades of pixels inside each triangle.
As mentioned above, conventional microprocessor based systems employing 3D graphics processing have experienced bandwidth limitations. For example, a microprocessor, such as X-86 is coupled to a 3D graphics chip via a PCI bus. An external memory stores information relating to the 3D model. The microprocessor performs the geometry and lighting calculations and transfers the results, which are information relating to the nodes of each triangle, via the PCI bus to the 3D graphics chip.
The 3D graphics chip includes a slope calculator that measures the slope of each side of the triangle. An interpolator calculates the shading colors of each pixel within a triangle based on the measured slopes. A texturing unit measures the textures of each pixel within the triangle based on the measured slopes and on the information stored in the texture map.
A separate frame buffer memory is employed to store the texture maps described above. Each texture map corresponds to the texture of an element used in the image. Furthermore, the frame buffer memory includes a separate buffer space referred to as a Z-buffer. The Z-buffer is employed to remove the hidden parts of a triangle when it is not intended to be displayed. Thus, when a plurality of objects are overlapped, invisible planes need to be removed in order to determine which edges and which planes of which objects are visible and display only the visible planes. Conventionally, various algorithms are employed for removing invisible planes as described in Fundamentals of Interactive Computer Graphics, J. D. Foley and A. Vandam (Addisson Wesley 1982), and incorporated herein by reference.
The Z-buffer stores, a Z-value, or a depth value of each pixel that needs to be displayed on a screen. Then, a Z-value of each point having x, y coordinate values, within a triangle is calculated and the obtained calculation result is compared with the Z-value corresponding to x,y coordinates. When the Z-value of a point is larger than the stored Z-value that point is considered to be hidden.
Thus, the microprocessor based system described above divides the graphics processing functions between the microprocessor and the 3D graphics chip. The microprocessor does the geometry and lighting steps, and provides triangle data to the graphics chip via the PCI bus. A typical graphics processing operation requires the processing of 1M triangles/sec. Each triangle contains about 50–60 bytes of information. This information includes, the x,y,z coordinates of each node, the color values, R,G, B, alpha, the texture coordinate values, u,v for each of the three nodes. Thus when each coordinate and color and texture value is represented by 4 bytes of information, the three nodes of each triangle may be defined by 108 (36×3) bytes of information This relates to a data transfer of 108 Mbytes/sec from the microprocessor to the 3D graphics chip. Thus, the PCI bus may experience severe bottle necking.
Another bandwidth limitation in implementing 3D graphics processing is data transfers from frame buffers to the 3D graphics chip. Usually, a typical model space may include 2–4 objects overlapping in each area. Thus, shading and texturing may be done 2–4 times in conjunction with Z buffering. For a 60 frame per second display, the data transfer rate between the frame buffer and 3D graphics chip is about 425 M bytes/sec (3 bytes/pixel×1024 pixel/line×768 lines×3 (shadings)×60 frames/sec), without the Z-buffering. For Z-buffering, this transfer rate is 520 Mbyte/sec (2 byte/pixel (Z-buffer read)×3×1024×768×60 frames/sec+(3+2) byte/pixel×1024×768×60) because of the read and write operation involved in Z-buffering. Texel fetching also requires a data transfer rate of 360 Mbytes/sec. Such data transfer rates are not feasible with the current memory technology. Thus, current 3D graphics arrangements employ substantially lower resolutions, which would not lead to realistic images.
Thus, there is a need to reduce the bandwidth delays associated with transfer of data from a microprocessor to a 3D graphics chip and bandwidth delays associated with transfer of data from a frame buffer to the 3D graphics chip.
Another disadvantage with prior art multimedia systems is their data transfer arrangement methods. In many data processing chip sets data is transferred from one or many processors to memory devices and input/output, I/O, subsystems, or other chip components known as functional units, via an appropriate bus structure. Typically, the bus structure includes a processor bus, a system bus and a memory bus. Thus, when there is a memory operation wherein data is required to be moved to or from a memory location to a processor, the system bus would cease to operate until the data movement from the memory location to the processor is completed. Similarly, when there is a data movement from an external device to a memory location, the processor bus would cease to operate until the data is moved to its intended location.
In order to alleviate the under utilization of bus subsystems as described above, U.S. Pat. No. 5,668,965 issued on Sep. 16, 1997, teaches the use of a controller that forms a three-way connection of three kinds of buses including a processor bus linked to at least one processor, a memory bus connected to a main memory, and a system bus linked to at least one connected device such as an input/output, I/O, device, thereby establishing interconnections between various buses. The controller includes data path switch means for transferring control signals and addresses through the control and address buses respectively of the three kinds of buses, and for generating a data path control signal to be supplied to the data switch means.
This arrangement allows the use of the buses on an independent basis. For example, when a processor on the processor bus conducts a processor/main memory access to access the main memory on the memory bus, data is transferred only via the processor and memory buses, allowing the system bus to operate independently.
However, the arrangement disclosed in the '965 patent does not provide for a priority based data movement. Furthermore, it does not disclose a mechanism to handle data transfers between endpoints that exhibit mismatched bandwidth requirements.
Additionally, conventional data movement arrangements have failed to address application-specific requirements. For example, when a data processor is employed for handling graphical images and displaying them on a screen, considerable throughput efficiency may be gained by taking into account the memory address patterns that are inherent with such graphical images.
Another disadvantage with conventional systems is that the resources employed by the data movement arrangements cannot be flexibly specified based on a corresponding data transfer between two end points. For example, some data movement arrangements employ fixed buffers to accommodate separate input/output, I/O, data transfers.
Thus, there is a need for a multimedia system that employs a data movement arrangement that overcomes the disadvantages discussed above, and specifically accommodates data transfers for an integrated media processor chip set that contains various system components such as processors, data cache, three dimensional graphics units, memory and input/output devices.